摘要
针对现有燃油估计方法大多只考虑飞机纵剖面、巡航阶段且模型参数复杂等问题,传统方法工程应用较差,提出一种采用T-S模糊神经网络的飞机燃油消耗估计模型。通过提取特定航班的若干次QAR数据及飞机性能,不同航段关键影响因素,实行航路分阶段建模。仿真中前件网络参数采用改进的衰减记忆全局关联函数和学习率自适应调整相结合的梯度下降算法进行计算,后件网络采用改进的遗传算法。实验结果表明,改进方法有效的提高了估计精度,更加契合工程应用需求,具有较高的实用价值。
Concerning that most existing fuel estimate model methods only consider plane profile, cruise phase, complex calculation and poor engineering application, based on T-S fuzzy neural network, this paper presented an estimation model for fuel consumption. The key parameters were extracted from several actual flight QAR data and aircraft performances of specific flight as the input training network. According to different key influencing factors of different legs, airway was modeled in stages. For the antecedent network parameter, we used an improvement gradient descent algorithm combined with fading memory global correlation function and adaptive adjustment; for consequent network, we used an improvement genetic algorithm. The simulation results show that the precision and engineering application are obviously improved for the nonlinear fuel estimation problem by this algorithm.
出处
《计算机仿真》
CSCD
北大核心
2015年第5期33-36,共4页
Computer Simulation
基金
国家自然科学基金(61079009)
中央高校基本科研业务费资助项目(3122013SY24)
中央高校基本科研业务费资助项目(3122013SY22)
关键词
模糊神经网络
燃油消耗
燃油估计
遗传算法
Fuzzy neural network
Fuel consumption
Fuel estimate
Genetic algorithms
